Abasy Atlas provides a comprehensive atlas of annotated functional systems (hereinafter also referred as modules), global network properties, and system-level elements predicted by the natural decomposition approach[1,2,3] (NDA) for reconstructed and meta-curated regulatory networks across a large range of bacteria, including pathogenic and biotechnologically relevant organisms. Regulatory networks could be thought of as a “take decisions” organ in bacteria. They not only sense stimulus and respond accordingly but, for example, in a complex environment, they “take composite decisions” to prioritize the assimilation and catabolism of carbon sources according to the metabolic preferences of each organism. To accomplish this, RNs composed by thousands of regulatory interactions, must follow well-defined organization principles governing their dynamics.
In the last decades of the 20th century, the first levels of gene organization were unveiled as the operon and the regulon. Currently, a few databases (RegulonDB, SubtiWiki, DBTBS, CoryneRegNet, and RegTransBase) extract, by manual curation of literature, the molecular knowledge about gene regulation in different organisms providing an invaluable source of information. Nevertheless, information on these databases never goes beyond the regulon level, whereas cumulative evidence has showed that regulatory networks are complex hierarchical-modular networks whose organizational and evolutionary principles are pivotal for determining the dynamics of the cell and still challenging our understanding. In this atlas, we take the first step towards a global understanding of the regulatory networks organization by building a cartography of the functional architecture for each of the best-studied organisms.
If you want to reference Abasy Atlas, the provided data or any generated image, please cite:
The NDA defines criteria to identify systems and system-level elements in a regulatory network, and rules to reveal its functional architecture by controlled decomposition (Figure 1). This biologically motivated approach mathematically derives the architecture and system-level elements from the global structural properties of a given regulatory network. It is based on two biological premises:
A module is a set of genes cooperating to carry out a particular physiological function, thus conferring different phenotypic traits to the cell.
Given the pleiotropic effect of global regulators, they must not belong to modules but rather coordinate them in response to general-interest environmental cues.
According to this approach, every gene in a regulatory network is predicted to belong to one out of four possible classes of system-level elements, which interrelate in a non-pyramidal, three-tier, hierarchy shaping the functional architecture as follows[1,2,3]: (1) Global regulators are responsible for coordinating both the (2) basal cell machinery, composed of exclusively globally regulated genes (EGRGs), and (3) locally autonomous modules (shaped by modular genes), whereas (4) intermodular genes integrate, at promoter level, physiologically disparate module responses eliciting a combinatorial processing of environmental cues (Figure 2).
There are a couple of advantages resulting from developing an ad-hoc method based on biological knowledge: (1) Global regulators (hubs) does not belong to any module. This is biologically important because they are not related to a particular physiologic function. (2) We identify that there are overlap among modules and it is mediated by the intermodular genes. None of this key features existing in bacterial regulatory networks could be identified with any of the methods commonly used to identify communities in complex networks.
Estimating regulatory networks completeness by leveraging their constrained complexity
The ability to quantify the total number of interactions in the regulatory network of an organism is a valuable insight that will leverage the daunting task of curation, prediction, and validation by enabling the inclusion of prior information about the network structure. But poor efforts have been directed towards the longstanding problem of how to assess the completeness of these networks. Traditionally, genomic coverage has been used as a proxy of completeness. The genomic coverage of a regulatory network is the fraction of genes in the network relative to the genome size. Nevertheless, this measure poses potential biases as it neglects regulatory redundancy and the combinatorial nature of gene regulation, thus potentially overestimating network completeness.
The addition of a global regulon or sigmulon (perhaps discovered by high-throughput methodologies) to a quite incomplete regulatory network could bias the genomic coverage. For example, assume you have a regulatory network with a genomic coverage of 15% (600 / 4,000) and 700 regulatory interactions. You then found a paper reporting the promoter mapping for the corresponding housekeeping sigma factor, whose sigmulon have 3,000 genes. Next, you integrate all these 3,000 new interactions to your original network to found that now your network have a new genomic coverage of ∼79% (3,150 / 4,000) and 3,650 interactions. The new high genomic coverage may suggests a highly complete network but in fact it is the same quite incomplete original network plus a single global sigmulon. To clarify this, assume that the total number of interactions is 10,105, then the completeness of this new network is ∼36% (3,650 / 10,105). Whereas the curation of a single housekeeping sigmulon increased the completeness ∼30%, the new completeness is low yet and the genomic coverage is highly overestimating when is used as a proxy for completeness.
Therefore, to correctly quantify the completeness of a regulatory network the total number of regulatory interactions is required. A recently found constrain in the complexity of regulatory networks allowed to estimate the total number of interactions a regulatory network has. Using all the regulatory networks available in Abasy Atlas, it was found that the density as function of the number of genes follows a power law as d ∼ n-α with α ≈ 1 (Figure 3), which is in agreement with the May-Wigner stability theorem. Network density is the fraction of potential interactions that are real interactions, thus a constraint in network density implies a constraint in the total number of interactions. It was found that the number of interactions also follows a power law as I ∼ nγ with γ = 2 - α (R2 = 0.98, Figure 4). This power law may be used to compute the total number of interactions (Itotal) in the regulatory network of an organism as Itotal ∼ (genome size)2 - α.
Disclaimer: Please note that original data contained in Abasy Atlas may be subject to rights claimed by third parties.
It is the responsibility of users of Abasy Atlas to ensure that their exploitation of the data does not infringe any of the rights of such third parties.